Open Question Answering on MATH500 (test)
0.94AccuracyGRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPO2026.02 | 0.94 | 3,962 | |
| FlashThink2026.02 | 0.938 | 3,050 | |
| AdaptThink2026.02 | 0.938 | 2,130 | |
| ESTARoptimization=RL2026.02 | 0.938 | 635 | |
| ESTAR-FTmode=fine-tuned2026.02 | 0.934 | 2,401 | |
| ESTAR-LITEearly_stopping=classifier-based2026.02 | 0.932 | 2,019 | |
| Length-Penaltyexplicit_length_penalty=true2026.02 | 0.924 | 3,190 | |
| O1-Pruner2026.02 | 0.916 | 2,856 | |
| No-Thinkingreasoning=disabled2026.02 | 0.85 | 1,139 |